CN116665763A - Metabolism path deducing method based on multi-view multi-tag learning - Google Patents

Metabolism path deducing method based on multi-view multi-tag learning Download PDF

Info

Publication number
CN116665763A
CN116665763A CN202310562934.3A CN202310562934A CN116665763A CN 116665763 A CN116665763 A CN 116665763A CN 202310562934 A CN202310562934 A CN 202310562934A CN 116665763 A CN116665763 A CN 116665763A
Authority
CN
China
Prior art keywords
molecular
representation
fingerprints
learning
diagram
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202310562934.3A
Other languages
Chinese (zh)
Other versions
CN116665763B (en
Inventor
郭菲
刘晓依
杨洪鹏
艾成伟
唐继军
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central South University
Original Assignee
Central South University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central South University filed Critical Central South University
Priority to CN202310562934.3A priority Critical patent/CN116665763B/en
Publication of CN116665763A publication Critical patent/CN116665763A/en
Application granted granted Critical
Publication of CN116665763B publication Critical patent/CN116665763B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16B—BIOINFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR GENETIC OR PROTEIN-RELATED DATA PROCESSING IN COMPUTATIONAL MOLECULAR BIOLOGY
    • G16B5/00—ICT specially adapted for modelling or simulations in systems biology, e.g. gene-regulatory networks, protein interaction networks or metabolic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/30—Prediction of properties of chemical compounds, compositions or mixtures
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16C—COMPUTATIONAL CHEMISTRY; CHEMOINFORMATICS; COMPUTATIONAL MATERIALS SCIENCE
    • G16C20/00—Chemoinformatics, i.e. ICT specially adapted for the handling of physicochemical or structural data of chemical particles, elements, compounds or mixtures
    • G16C20/70—Machine learning, data mining or chemometrics
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Abstract

The application relates to the technical field of metabonomics, and discloses a metabolic pathway inference method based on multi-view and multi-tag learning, which comprises the following steps: step one: first, the SMILES sequence is decomposed into atomic scale tags, then each tag is mapped to a vector with a fixed dimension, after which these embedded representations are passed to the bi-directional LSTM, and finally a sequence-based molecular representation is obtained by the MLP layer, step two: the original molecular topology was obtained by using RDKit. The method comprises the steps of splicing four complementary fingerprints, then encoding the spliced molecular fingerprints through a multi-layer perceptron to obtain a representation based on the molecular fingerprints, combining the four different types of molecular fingerprints with a simplified molecular input line entry system representation to form a mixed fingerprint, and inputting the mixed fingerprint vector into the MLP with a nonlinear activation function, so that compared with the prior art, the method has obvious advantages, and a more accurate and effective solution is provided for predicting a path set of a compound.

Description

Metabolism path deducing method based on multi-view multi-tag learning
Technical Field
The application relates to the technical field of metabonomics, in particular to a metabolic pathway inference method based on multi-view and multi-tag learning.
Background
In the field of metabolic pathway multi-label prediction, existing implementation schemes mainly comprise a network-based method and a graph neural network-based method. Network-based methods, such as chemical-chemical interaction networks, protein-protein interaction networks, and chemical-protein interaction networks, utilize biological network information to predict metabolic pathways in which a compound participates. However, these methods may ignore isolated compounds in the network when predicting metabolic pathways by integrating multiple layers of biological networks. On the other hand, graph neural network-based methods, such as graph rolling networks (GCN) and graph annotation force networks (GATs), have been successfully applied to extract graph structures and semantic features to implement graph rolling operations of molecular structures.
The existing multi-tag metabolic pathway prediction method has some limitations, and mainly comprises the following steps: firstly, they are typically based on a single view-angle representation of data, which may not adequately capture the diversity and complexity of the compounds; secondly, these methods focus on compound structural information, but ignore multiple different feature sets of the compound, such as SMILES, fingerprints, and maps; third, current methods have limited generalization performance because they use only one learnable convolution kernel when learning the molecular structure representation, which may affect the prediction accuracy of the model.
Disclosure of Invention
Aiming at the defects of the prior art, the application provides a metabolic pathway inference method based on multi-view and multi-tag learning, which solves the problem that the diversity and complexity of the compounds cannot be fully captured.
In order to achieve the above purpose, the application is realized by the following technical scheme: a metabolic pathway inference method based on multi-view multi-tag learning, comprising the steps of:
step one: the SMILES sequence is first decomposed into atomic scale tags, each tag is then mapped to a vector of fixed dimensions, after which these embedded representations are passed to the bi-directional LSTM, and finally the sequence-based feature representations are obtained by the MLP layer.
Step two: the method comprises the steps of obtaining an original molecular topological structure by using an RDkit, taking the original molecular topological structure diagram as diagram data of a GCN layer, learning embedded representation of nodes in the diagram by the GCN, then utilizing Set2Set 4 global pooling operation of an attention mechanism based on iterative content to aggregate embedded vectors of the nodes in the diagram, and finally obtaining graphic embedded representation of molecules by an MLP layer.
Step three: and splicing four complementary fingerprints, encoding the spliced molecular fingerprints through a multi-layer perceptron to obtain a representation based on the molecular fingerprints, combining the four different types of molecular fingerprints with a simplified molecular input line entry system representation to form a mixed fingerprint, and inputting the mixed fingerprint vector into the MLP with a nonlinear activation function.
Step four: by calculating the Q (Query), K (Key) and V (Value) matrices, the model can assign weights to the representations of different views, and meanwhile, a multi-head attention mechanism is introduced, so that the model can pay attention to information from different views at the same time, and finally, the output of the multi-head attention is connected and input to an MLP layer to obtain the prediction score of each metabolic pathway.
Preferably, the plurality of molecules in the second step are a molecular sequence, topology information and a fingerprint.
Preferably, a metabolic pathway inference system based on multi-view multi-tag learning, comprising:
the sequence representation learning module is used for learning SMILES sequence information of molecules by using Bi-directional LSTM (Bi-LSTM) so as to capture original information of molecular structures.
The graph representation learning module is used for learning topology representation information of the molecular graph by using a graph rolling network (GCN), and adopts a Set2Set [4] global pooling operation of an attention mechanism based on iterative content to aggregate embedded vectors of nodes in the graph.
The fingerprint representation learning module is used for splicing four complementary fingerprints, and then encoding the spliced molecular fingerprints through the multi-layer perceptron to obtain the representation based on the molecular fingerprints.
An attention-based fusion module for efficiently integrating multi-view features of molecules.
Working principle: the method comprises the steps of firstly decomposing an SMILES sequence into atomic level marks, mapping each mark to a vector with fixed dimension, transferring the embedded marks to a bidirectional LSTM, finally obtaining a characteristic representation based on the sequence through an MLP layer, obtaining an original molecular topological structure through using RDkit, then taking an original molecular topological structure diagram as image data of a GCN layer, aggregating embedded vectors of nodes in the image through a Set2Set [4] global pooling operation based on an attention mechanism of iterative content, finally obtaining a graphic embedded representation of molecules through one MLP layer, splicing four complementary fingerprints, encoding the spliced molecular fingerprints through a multi-layer perceptron, obtaining a representation based on the molecular fingerprints, combining the four different types of molecular fingerprints with a simplified molecular input line entry system representation to form a mixed fingerprint, inputting the mixed fingerprint vector into a multi-point model with a nonlinear activation function, and simultaneously distributing weights for different representations through calculating Q (Query), K (Key) and V (Value) matrixes, simultaneously inputting the model into the multi-point model, and simultaneously inputting the multi-point visual angle information from the multi-point visual angle model, and simultaneously inputting the multi-point visual angle model to the multi-point visual angle model, and outputting the multi-point visual angle model from the multi-point visual angle.
The application provides a metabolic pathway inference method based on multi-view and multi-tag learning. The beneficial effects are as follows:
according to the application, multiple molecular representations including molecular sequences, fingerprints and topology information can be integrated by adopting the attention-based multi-view fusion network, so that the structure and attribute information of the molecules are fully captured, the model can effectively learn and mine the correlation between different representations by introducing a multi-head attention mechanism, the prediction performance of the model is further enhanced, multiple molecular representations are learned by utilizing a Bi-directional long-short-time memory (Bi-LSTM) structure, a fingerprint neural network (FP-NN) and a graph rolling network (GCN), rich molecular feature representations are generated, the prediction capability of the model in complex tasks is improved, and compared with the prior art, the multi-view fusion, multi-head attention mechanism, multiple feature learning methods, multi-label learning and the like, the method has obvious advantages, and a more accurate and effective solution is provided for predicting a path set of a compound.
Drawings
FIG. 1 is a complete flow chart of the present application;
FIG. 2 is a technical roadmap of the application;
FIG. 3 is a graph comparing different models of the data of the present application;
FIG. 4 is an experimental view of ablation of a model of the present application;
FIG. 5 is a graph of the performance evaluation of each class of metabolic pathways according to the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
Embodiment one:
referring to fig. 1-5, an embodiment of the present application provides a metabolic pathway inference method based on multi-view and multi-tag learning, including the following steps:
step one: the SMILES sequence is first decomposed into atomic scale tags, each tag is then mapped to a vector of fixed dimensions, after which these embedded representations are passed to the bi-directional LSTM, and finally the sequence-based feature representations are obtained by the MLP layer.
Step two: the method comprises the steps of obtaining an original molecular topological structure by using an RDkit, taking the original molecular topological structure diagram as diagram data of a GCN layer, learning embedded representation of nodes in the diagram by the GCN, then utilizing Set2Set 4 global pooling operation of an attention mechanism based on iterative content to aggregate embedded vectors of the nodes in the diagram, and finally obtaining graphic embedded representation of molecules by an MLP layer.
Step three: and splicing four complementary fingerprints, encoding the spliced molecular fingerprints through a multi-layer perceptron to obtain a representation based on the molecular fingerprints, combining the four different types of molecular fingerprints with a simplified molecular input line entry system representation to form a mixed fingerprint, and inputting the mixed fingerprint vector into the MLP with a nonlinear activation function.
Step four: by calculating the Q (Query), K (Key) and V (Value) matrices, the model can assign weights to the representations of different views, and meanwhile, a multi-head attention mechanism is introduced, so that the model can pay attention to information from different views at the same time, and finally, the output of the multi-head attention is connected and input to an MLP layer to obtain the prediction score of each metabolic pathway.
And in the second step, the various molecules are a molecular sequence, topology information and fingerprints.
Embodiment two:
a metabolic pathway inference system based on multi-view, multi-tag learning, comprising:
the sequence representation learning module is used for learning SMILES sequence information of molecules by using Bi-directional LSTM (Bi-LSTM) so as to capture original information of molecular structures.
The graph representation learning module is used for learning topology representation information of the molecular graph by using a graph rolling network (GCN), and adopts a Set2Set [4] global pooling operation of an attention mechanism based on iterative content to aggregate embedded vectors of nodes in the graph.
The fingerprint representation learning module is used for splicing four complementary fingerprints, and then encoding the spliced molecular fingerprints through the multi-layer perceptron to obtain the representation based on the molecular fingerprints.
An attention-based fusion module for efficiently integrating multi-view features of molecules.
Embodiment III:
referring to fig. 1, the present application utilizes three different feature sets, module a, namely SMILES, molecular graph and fingerprint. Integrating these different data views allows the model to better represent the features and relationships between compounds and pathways, and module B integrates multiple composite encoders, each specifically designed to capture a different aspect of the chemical structure, in order to further enhance learning of multiple different feature data sets. Module C, a fusion module for the model, effectively fuses information from different views of the compound using an attention-based mechanism. This enables the model to identify and combine the most relevant information in each view. Finally, module D measures the proximity between the learning token and the metabolic pathway for the pathway predictor to determine pathway participation.
Embodiment four:
referring to FIG. 3, the present application predicts a comparison of F1 score, AUC, AUPR, recall and Precision using different methods on 11 different metabolic pathway datasets by 4129 metabolites. Comparing the present application (MVML-MPI) with a graph-roll-up network based model, and a graph-attention network based model with an autopsr, it can be seen that the predicted F1 score, AUC, AUPR, recall and Precision of the present application all achieve the best values.
Fifth embodiment:
referring to fig. 4, the present application compares the contributions of different modules to the prediction performance, and results show that the use of a fusion module based on an attention mechanism can improve the prediction performance, and is superior to a model using only a single molecular feature view, and the present application (MVML-MPI) model using a SMILES view performs best in different molecular encoders and fusion modules, and can improve the accuracy, precision, recall, and F1 score. Furthermore, the MVML-MPI model has advantages in predicting the multi-tag metabolic pathway compared to a model based on simple concatenation of feature sets only, emphasizing the key role of attention-mechanism based fusion strategies in multi-tag metabolic pathway prediction.
Example six:
referring to fig. 5, the present application further evaluates the ability of the present application to access tags in underrepresented access categories by classification performance evaluation of each type of metabolic access, using accuracy, F1 score, recall and precision as compared to the AutoMSR model.
Although embodiments of the present application have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the application, the scope of which is defined in the appended claims and their equivalents.

Claims (3)

1. A metabolic pathway inference method based on multi-view and multi-tag learning, comprising the steps of:
step one: the SMILES sequence is first decomposed into atomic scale tags, each tag is then mapped to a vector of fixed dimensions, after which these embedded representations are passed to the bi-directional LSTM, and finally the sequence-based feature representations are obtained by the MLP layer.
Step two: the method comprises the steps of obtaining an original molecular topological structure by using an RDkit, taking the original molecular topological structure diagram as diagram data of a GCN layer, learning embedded representation of nodes in the diagram by the GCN, then utilizing Set2Set 4 global pooling operation of an attention mechanism based on iterative content to aggregate embedded vectors of the nodes in the diagram, and finally obtaining graphic embedded representation of molecules by an MLP layer.
Step three: and splicing four complementary fingerprints, encoding the spliced molecular fingerprints through a multi-layer perceptron to obtain a representation based on the molecular fingerprints, combining the four different types of molecular fingerprints with a simplified molecular input line entry system representation to form a mixed fingerprint, and inputting the mixed fingerprint vector into the MLP with a nonlinear activation function.
Step four: by calculating the Query, key and Value matrix, the model can allocate weights for the representation of different view angles, meanwhile, a multi-head attention mechanism is introduced, so that the model can pay attention to information from different view angles at the same time, and finally, the output of the multi-head attention is connected and input to an MLP layer to obtain the prediction score of each metabolic pathway.
2. The metabolic pathway inference method based on multi-view and multi-tag learning according to claim 1, wherein the molecules in the first, second and third steps are a molecular sequence, topology information and a fingerprint.
3. A metabolic pathway inference system based on multi-view multi-tag learning according to claim 1 or 2, characterized by comprising:
the sequence representation learning module is used for learning the SMILES sequence information of the molecule by utilizing the bidirectional LSTM so as to capture the original information of the molecular structure.
The diagram representation learning module is used for learning topology representation information of the molecular diagram by using the diagram convolution network and aggregating embedded vectors of nodes in the diagram by adopting a Set2Set global pooling operation based on an attention mechanism of iterative content.
The fingerprint representation learning module is used for splicing four complementary fingerprints, and then encoding the spliced molecular fingerprints through the multi-layer perceptron to obtain the representation based on the molecular fingerprints.
An attention-based fusion module for efficiently integrating multi-view features of molecules.
CN202310562934.3A 2023-05-18 2023-05-18 Metabolism path deducing method based on multi-view multi-tag learning Active CN116665763B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310562934.3A CN116665763B (en) 2023-05-18 2023-05-18 Metabolism path deducing method based on multi-view multi-tag learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310562934.3A CN116665763B (en) 2023-05-18 2023-05-18 Metabolism path deducing method based on multi-view multi-tag learning

Publications (2)

Publication Number Publication Date
CN116665763A true CN116665763A (en) 2023-08-29
CN116665763B CN116665763B (en) 2024-05-03

Family

ID=87725228

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310562934.3A Active CN116665763B (en) 2023-05-18 2023-05-18 Metabolism path deducing method based on multi-view multi-tag learning

Country Status (1)

Country Link
CN (1) CN116665763B (en)

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013181758A1 (en) * 2012-06-05 2013-12-12 Mcmaster University Screening method and systems utilizing mass spectral fragmentation patterns
US20190272890A1 (en) * 2017-07-25 2019-09-05 Insilico Medicine, Inc. Deep proteome markers of human biological aging and methods of determining a biological aging clock
CN111223527A (en) * 2015-12-07 2020-06-02 齐默尔根公司 Improvement of microbial strains by using HTP genome engineering platform
US20200176087A1 (en) * 2018-12-03 2020-06-04 Battelle Memorial Institute Method for simultaneous characterization and expansion of reference libraries for small molecule identification
CN112435715A (en) * 2020-11-12 2021-03-02 中国科学院深圳先进技术研究院 Metabolic path prediction method and device, terminal device and storage medium
CN112562869A (en) * 2021-02-24 2021-03-26 北京中医药大学东直门医院 Drug combination safety evaluation system, method and device
US20210158903A1 (en) * 2019-11-21 2021-05-27 Hitachi, Ltd. Information processing system and search method
CN113066526A (en) * 2021-04-08 2021-07-02 北京大学 Hypergraph-based drug-target-disease interaction prediction method
CN114360659A (en) * 2021-12-21 2022-04-15 武汉大学 Biological reverse synthesis method and system combining and-or tree and single-step reaction rule prediction
US20220165359A1 (en) * 2020-11-23 2022-05-26 Peptilogics, Inc. Generating anti-infective design spaces for selecting drug candidates
CN114898815A (en) * 2022-05-10 2022-08-12 清华大学深圳国际研究生院 Homogeneous interaction prediction method and device based on spatial structure in field of drug discovery
CN114927173A (en) * 2022-04-06 2022-08-19 西北工业大学 Metabolic path prediction method based on label correlation and graph representation learning

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2013181758A1 (en) * 2012-06-05 2013-12-12 Mcmaster University Screening method and systems utilizing mass spectral fragmentation patterns
CN111223527A (en) * 2015-12-07 2020-06-02 齐默尔根公司 Improvement of microbial strains by using HTP genome engineering platform
US20190272890A1 (en) * 2017-07-25 2019-09-05 Insilico Medicine, Inc. Deep proteome markers of human biological aging and methods of determining a biological aging clock
US20200176087A1 (en) * 2018-12-03 2020-06-04 Battelle Memorial Institute Method for simultaneous characterization and expansion of reference libraries for small molecule identification
US20210158903A1 (en) * 2019-11-21 2021-05-27 Hitachi, Ltd. Information processing system and search method
CN112435715A (en) * 2020-11-12 2021-03-02 中国科学院深圳先进技术研究院 Metabolic path prediction method and device, terminal device and storage medium
US20220165359A1 (en) * 2020-11-23 2022-05-26 Peptilogics, Inc. Generating anti-infective design spaces for selecting drug candidates
CN112562869A (en) * 2021-02-24 2021-03-26 北京中医药大学东直门医院 Drug combination safety evaluation system, method and device
CN113066526A (en) * 2021-04-08 2021-07-02 北京大学 Hypergraph-based drug-target-disease interaction prediction method
CN114360659A (en) * 2021-12-21 2022-04-15 武汉大学 Biological reverse synthesis method and system combining and-or tree and single-step reaction rule prediction
CN114927173A (en) * 2022-04-06 2022-08-19 西北工业大学 Metabolic path prediction method based on label correlation and graph representation learning
CN114898815A (en) * 2022-05-10 2022-08-12 清华大学深圳国际研究生院 Homogeneous interaction prediction method and device based on spatial structure in field of drug discovery

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
PAOLA RUIZ PUENTES 等: "PharmaNet: Pharmaceutical discovery with deep recurrent neural networks", 《PLOS ONE》, vol. 16, no. 4, 26 April 2021 (2021-04-26) *
赵其昌: "基于深度注意模型的药物蛋白质关系预测", 《中国优秀硕士学位论文全文数据库 (医药卫生科技辑)》, no. 2, 15 February 2021 (2021-02-15) *

Also Published As

Publication number Publication date
CN116665763B (en) 2024-05-03

Similar Documents

Publication Publication Date Title
CN111400620B (en) User trajectory position prediction method based on space-time embedded Self-orientation
US11176446B2 (en) Compositional prototypes for scalable neurosynaptic networks
Xu et al. Predicting destinations by a deep learning based approach
CN112419710A (en) Traffic congestion data prediction method, traffic congestion data prediction device, computer equipment and storage medium
CN114493014A (en) Multivariate time series prediction method, multivariate time series prediction system, computer product and storage medium
Zuo et al. Graph convolutional networks for traffic forecasting with missing values
CN116126341A (en) Model compiling method, device, computer equipment and computer readable storage medium
Liu et al. Road segmentation with image-LiDAR data fusion in deep neural network
Rao et al. Sliding space-disparity transformer for stereo matching
CN116665763B (en) Metabolism path deducing method based on multi-view multi-tag learning
CN113704276A (en) Map updating method and device, electronic equipment and computer readable storage medium
Bi et al. Critical direction projection networks for few-shot learning
CN116680456A (en) User preference prediction method based on graph neural network session recommendation system
Che et al. Pedestrian multiple-object tracking based on FairMOT and circle loss
CN114462707B (en) Web service multidimensional QoS joint prediction method based on feature depth fusion
CN115203412A (en) Emotion viewpoint information analysis method and device, storage medium and electronic equipment
CN114418189A (en) Water quality grade prediction method, system, terminal device and storage medium
Liu et al. Analysis of the mixed teaching of college physical education based on the health big data and blockchain technology
Senuma et al. GEAR: a graph edge attention routing algorithm solving combinatorial optimization problem with graph edge cost
Zhong et al. Densely packed object detection with transformer-based head and EM-merger
Tarakanov et al. Identification of cellular automata by immunocomputing.
CN112466410B (en) Method and device for predicting binding free energy of protein and ligand molecule
CN116883914B (en) Video segmentation method and device integrating semi-supervised and contrast learning
CN115908119B (en) Face image beautifying processing method and system based on artificial intelligence
Wang et al. Keymemoryrnn: A flexible prediction framework for spatiotemporal prediction networks

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant